Python Notes15:hdf5存储图像文件

1. 图片存为hdf5（灰度）

1.1 写数据

• 将文件夹下的图片写成hdf5文件存储

import os
import h5py
import cv2
import numpy as np

# 中文路径读不出来
img_path = r'D:\DataDemo'
out_h5_path = r'D:\DataDemo\test_hdf5.h5'

# 遍历指定地址下的所有图片,💥尺寸大小一样💥的的jpg文件
img_list = sorted(os.listdir(img_path))

# 确认图片的大小
flag = 0  # lags=1 读取彩色图像(BGR) ，flags=0 读取为灰度图像
image = cv2.imread(img_path + '\\2.jpg', flags=flag)
# print(image)

shape = np.array(image, dtype=np.uint8).shape
# print(shape[0])
# print(shape[1])

num_img = len(img_list)

# dataset = np.zeros((num_img, shape[0], shape[1], 3), np.float)
dataset = np.zeros((num_img, shape[0], shape[1]), np.float)
# print(dataset)

cnt_num = 0
for img in img_list:
    if img.endswith(".jpg"):
        # print(img_path+'\\'+img)
        # flags=1 读取彩色图像(BGR)  # flags=0 读取为灰度图像
        gray_img = cv2.imread(img_path+'\\'+img, flags=flag)
        print(gray_img)
        # dataset[cnt_num, :, :, :] = gray_img
        dataset[cnt_num, :, :] = gray_img
        cnt_num += 1
# 存储为npy
np.save(img_path+'\\gray.npy', dataset)
# print('检验是否读取成功:')
# print(dataset[1])
with h5py.File(out_h5_path, 'w') as f:
    f['gray'] = dataset  # 将数据写入文件的主键train下面

1.2 读数据

import h5py
import cv2
from matplotlib import pyplot as plt

# 打开h5文件
filename = r'D:\DataDemo\test_hdf5.h5'
f = h5py.File(filename, 'r')
data = f['gray'][:]

# cv成图
cv2.imshow('data', data[1].astype("uint8"))
cv2.waitKey(0)
cv2.destroyAllWindows()

# plt成图，因为opencv读取的图片是BGR形式，想要用plt显示，则必须转成RGB形式
image = cv2.cvtColor(data[1].astype("uint8"), cv2.COLOR_BGR2RGB)
plt.imshow(image)
plt.xticks([]), plt.yticks([])
plt.show()

2. 图片存为hdf5（RGB）

2.1 写数据

• 将文件夹下的图片写成hdf5文件存储

import os
import h5py
import cv2
from matplotlib import pyplot as plt
import numpy as np

# 中文路径读不出来
img_path = r'D:\DataDemo'
out_h5_path = r'D:\DataDemo\test_hdf5.h5'

# 遍历指定地址下的所有图片,💥尺寸大小一样💥的的jpg文件
img_list = sorted(os.listdir(img_path))

# 确认图片的大小
flag = 1  # lags=1 读取彩色图像(BGR) ，flags=0 读取为灰度图像
image = cv2.imread(img_path + '\\2.jpg', flags=flag)

num_img = len(img_list)

# np.uint8符号整数（0 to 255）
shape = np.array(image, dtype=np.uint8).shape
# print(shape)

# RGB三原色 、创建四维数组
dataset = np.zeros((num_img, shape[0], shape[1], 3), np.float)
cnt_num = 0
for img in img_list:
    if img.endswith(".jpg"):
        # flags=1 读取彩色图像(BGR)  # flags=0 读取为灰度图像
        img_data = cv2.imread(img_path+'\\'+img, flags=flag)
        # print(img_data)
        dataset[cnt_num, :, :, :] = img_data
        cnt_num += 1

print(dataset[0] == image)
print(dataset[1] == image)
with h5py.File(out_h5_path, 'w') as f:
    f['train'] = dataset

2.2 读数据

• 将h5数据成图

'''
Description: henggao_note
version: v1.0.0
Date: 2022-04-21 16:27:06
LastEditors: henggao
LastEditTime: 2022-04-21 16:35:38
'''
import h5py
import cv2
from matplotlib import pyplot as plt
# 打开h5文件
filename = r'D:\DataDemo\test_hdf5.h5'
f = h5py.File(filename, 'r')
data = f['train'][:]
# print(data[1])

# img_path = r'D:\DataDemo\2.jpg'
# data1 = cv2.imread(img_path)
# cv2.imshow('data',data1)

# print(data[1] == data1)
# 矩阵A(numpy.ndarray)转换数值类型B = A.astype("uint8")，否则会报错
# data[i],显示第几张

# cv成图
# cv2.imshow('data', data[1].astype("uint8"))
# cv2.waitKey(0)
# cv2.destroyAllWindows()

# plt成图，因为opencv读取的图片是BGR形式，想要用plt显示，则必须转成RGB形式
image = cv2.cvtColor(data[1].astype("uint8"), cv2.COLOR_BGR2RGB)
plt.imshow(image)
plt.xticks([]), plt.yticks([])
plt.show()

3. 数据模型分析

3.1 灰度数据矩阵

[[106 127 128 ... 175 160 172]
 [ 29  41  89 ... 166 177 190]
 [ 68  23  18 ... 173 182 137]
 ...
 [ 23  30  15 ...  22  11   7]
 [ 15  28  24 ...  33  25   8]
 [ 20  24  36 ...  34  32   5]]

3.2 RGB数据举证

[[[154 113  74]
  [178 135  92]
  [186 137  87]
  ...
  [216 184 143]
  [204 168 128]
  [216 180 140]]

 [[ 79  37   0]
  [ 92  49   6]
  [144  98  51]
  ...
  [209 175 132]
  [221 185 145]
  [236 198 156]]

 [[119  77  30]
  [ 73  32   0]
  [ 64  26   0]
  ...
  [219 182 138]
  [230 190 148]
  [185 146 102]]

 ...

 [[ 39  26  12]
  [ 44  33  19]
  [ 27  18   5]
  ...
  [ 31  25  14]
  [ 18  13   4]
  [ 11  10   0]]

 [[ 33  18   2]
  [ 46  31  15]
  [ 40  27  13]
  ...
  [ 42  34  27]
  [ 32  27  18]
  [ 13  11   0]]

 [[ 39  23   6]
  [ 42  28  10]
  [ 54  38  25]
  ...
  [ 43  35  28]
  [ 39  34  25]
  [ 12   8   0]]]